Article pubs.acs.org/est
In Situ Spatially and Temporally Resolved Measurements of Salt Concentration between Charging Porous Electrodes for Desalination by Capacitive Deionization Matthew E. Suss,†,‡ P.M. Biesheuvel,§,⊥ Theodore F. Baumann,‡ Michael Stadermann,‡ and Juan G. Santiago*,† †
Department of Mechanical Engineering, Stanford University, 440 Escondido Mall, Stanford, California 94305, United States Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, California 94551, United States § Wetsus, Centre of Excellence for Sustainable Water Technology, Agora 1, 8934 CJ, Leeuwarden, The Netherlands ⊥ Department of Environmental Technology, Wageningen University, Bornse Weilanden 9, 6708 WG, Wageningen, The Netherlands ‡
S Supporting Information *
ABSTRACT: Capacitive deionization (CDI) is an emerging water desalination technique. In CDI, pairs of porous electrode capacitors are electrically charged to remove salt from brackish water present between the electrodes. We here present a novel experimental technique allowing measurement of spatially and temporally resolved salt concentration between the CDI electrodes. Our technique measures the local fluorescence intensity of a neutrally charged fluorescent probe which is collisionally quenched by chloride ions. To our knowledge, our system is the first to measure in situ and spatially resolved chloride concentration in a laboratory CDI cell. We here demonstrate good agreement between our dynamic measurements of salt concentration in a charging, millimeter-scale CDI system to the results of a modified Donnan porous electrode transport model. Further, we utilize our dynamic measurements to demonstrate that salt removal between our charging CDI electrodes occurs on a longer time scale than the capacitive charging time scales of our CDI cell. Compared to typical measurements of CDI system performance (namely, measurements of outflow ionic conductivity), our technique can enable more advanced and bettercontrolled studies of ion transport in CDI systems, which can potentially catalyze future performance improvements.
■
INTRODUCTION Capacitive deionization (CDI) is an emerging water desalination technique which uses a porous electrode capacitor to remove salt ions from brackish water.1 A typical CDI cell consists of two porous carbon electrodes separated by a porous separator element or an open channel. A voltage of roughly 1 V is applied across the porous electrodes, causing salt ions to electromigrate to the oppositely charged electrode and to be held electrostatically in electric double layers (EDLs) at the pore surface.2−4 The most common architecture of a CDI cell is a flow-between architecture, whereby the feedwater flows between the two charging porous electrodes.1,5−7 Alternative CDI architectures can use flow-through electrodes to enable faster and more efficient desalination,8 ion exchange membranes along electrode surfaces to improve electrode charge efficiency,9−11 or flow electrodes to enable desalination of high salinity feedwaters such as seawater.12 Once charged, the CDI cell must then be discharged to allow for subsequent desalination cycles.1 Discharging releases salt held in the EDLs, resulting in a brine solution.13 Since investigations into CDI-type systems began in the 1960s,14 the most widely used experimental technique to investigate CDI system desalination performance has been conductivity measurements of the system’s effluent.1,8,15−25 This typically involves a conductivity sensor placed in series © 2014 American Chemical Society
with, and downstream of, the CDI system. The latter technique is well suited for characterizing and evaluating the desalination performance of CDI cells; however, it is not as well suited for spatially resolved studies of ion transport in CDI systems. Notably, conductivity measurements can be affected by flow effects in the cell and the rest of the system components, such as mechanical or hydrodynamic dispersion, and typically cannot provide information on the contributions of individual electrodes. Classical electrochemical techniques such as cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) are often used to characterize CDI cell or individual CDI electrode capacitance,24,26,27 but the data from these techniques can be difficult to interpret for CDI cells performing desalination, as the desalination process is not reasonably represented by linear circuit models.28 Thus, a technique which can perform in situ (in a laboratory cell) measurements with spatial and temporal resolution of salt concentration in a charging CDI cell represents a significant advance in the experimental study and diagnostics of CDI cell performance. Such a technique can be used to study ion transport and Received: Revised: Accepted: Published: 2008
August 21, 2013 January 7, 2014 January 8, 2014 January 8, 2014 dx.doi.org/10.1021/es403682n | Environ. Sci. Technol. 2014, 48, 2008−2015
Environmental Science & Technology
Article
Figure 1. (a) Schematic of the capacitive deionization (CDI) cell used in this study. The cell consists of two 1 mm thick hierarchical carbon aerogel monolith (HCAM) electrodes which were held 1.5 mm apart. The HCAM electrodes have a hierarchical pore structure consisting of micrometerscale “macropores” and nanoscale “micropores”. (b) Schematic of a cross section of the setup used to visualize ion dynamics between two porous HCAM electrodes. The 1 mm thick electrodes were inserted into a transparent borosilicate glass channel and separated by a distance of 1.5 mm using a thin acrylic frame/spacer. The acrylic frame is three sided, with an open fourth side which permits UV light to enter into a 5 × 5 × 1.5 mm space between electrodes. The electrodes and space between them is filled with a KCl salt and SPQ probe solution. UV light excites the chloridesensing SPQ, whose fluorescent emissions are captured by our microscope objective and CCD camera.
dynamics in a cell without flow, whereas in future studies flow can be included, and its effects on the charging dynamics are studied separately.
desalination in the presence or absence of flow and can enable well-controlled, fundamental studies, the results of which can potentially guide performance enhancements. In this work, we demonstrate a simple and novel experimental system for in situ measurements of spatially and temporally resolved salt concentration between electrodes in an actively charging CDI cell. Our system measures the local fluorescence signal of a neutral (thus uncharged) probe molecule added to the background salt-water solution. This probe molecule’s fluorescence is quenched by collisions with chloride ions.29 As the probe is neutral, it does not electromigrate within our charging CDI cell, and we thus can simply correlate local fluorescence intensity between our charging electrodes to local chloride ion (salt) concentration. Our system is, to our knowledge, the first to demonstrate in situ temporally and spatially resolved measurements of chloride concentration in a CDI cell. Demirer et al. presented visualizations of charged dye species in a microfluidic CDI system with microfabricated pseudoporous electrodes.30 Their cell contained an aqueous solution with the charged (ionic) dye species at about 1 μM concentrations, and they correlated fluorescence intensity to local dye concentration.30 The latter technique was not used to measure the concentration of a nonfluorescent salt in a CDI cell (e.g., NaCl) and may not be well suited for such measurements (as charged dyes can themselves electromigrate). Roelofs et al. also performed visualizations of charged dyes in a microfluidic CDI-type system, and utilized a dye with pH dependent fluorescence to observe the onset of pH waves from planar electrodes.31 Sharma et al. utilized neutron imaging applied to a CDI cell to visualize the local concentration of gadolinium ions both within the pore space of mesoporous carbon electrodes and between the electrodes.32 In this work, we measure the local chloride concentration between the electrodes of a charging, realistically sized (millimeter to centimeter scale) CDI cell using a benchtop fluorescence microscope. We further compare our measurements to a modified Donnan (mD) CDI dynamic transport model.33 Importantly, we here study the desalination
■
THEORY In this section, we describe the dynamic transport model based on porous electrode theory, which we used to compare to our experimental data. We used a modified Donnan (mD) model for the electric double layer (EDL) structure, which has been used to model desalination in CDI systems using multiscale porous electrodes.33,34 The transport model assumes an electrode pore structure which is composed of larger transport pathways (interparticle pore space) located between micrometer-scale carbon particles (form this point onward, we term these pores “macropores”; to indicate features which have typically >50 nm characteristic scales) and the smaller intraparticle pores (which we will term “micropores”; and which typically have cmA initially. For the model calculations, we assumed a symmetric CDI cell, and thus, the applied cell voltage relates to the solid phase electrode potential (relative to the midplane), φ1, by: Vcell = 2φ1VT.
where ΔφD is the Donnan potential and Δφst is the potential drop across the Stern layer. Assuming equilibrium between the micro- and macropore, we can write the following relation for concentration of ion i in the micropore: cmi , i = cmA exp( −ziΔφD + μatt , i )
(8)
We also included in our model domain a separator layer between the porous electrodes, as is in our experimental cell (see Figure 1a). In the section of the model domain representing the separator (le ≤ x ≤ le+ls, where le is the electrode thickness and ls is the separator thickness), we removed the salt and charge sink terms in eqs 7 and 8 (terms led by the factor pmi). Thus, σmi is not a dependent variable in the space between electrodes, and we did not solve for eq 9 in this part of the domain. Further, in the separator, the parameter pmA in the model was replaced by the porosity of the separator layer, ps. For an open (nonporous) separator, ps can be set to unity. At the internal boundaries of our model domain representing the electrode−separator interface (x = le and ls + le), we ensured the continuity of cmA and φ, as well as imposed a matching condition for salt flux and of ionic current:
(3)
φ1 − φ = ΔφD + Δφst
∂σmi ∂φ ⎞ ∂ ⎛ ⎜2p Dc ⎟ = mA mA ⎝ ∂t ∂x ∂x ⎠
φ = φ1 + sinh−1(σmi /2cmAe μatt ) + σmiF /(Cst , volVT )
Here, cmA is the salt concentration in the electroneutral macropores, and cmi is the total concentration of ions in the micropores (cmi = cmi,a + cmi,c, where the subscripts a and c refer to anions and cations, respectively). These concentrations are quantities volume averaged over scales larger than the macropore topologic features but small compared to the electrode and cell dimensions. In the macropores (and between electrodes), the local concentration of anions equals that of the cations via the electroneutrality condition,36,37 and thus, cmA,a = cmA,c = cmA (and we can use the terms “ion concentration” and “salt concentration” interchangeably). In the micropore, anion and cation concentrations are not equal in order to balance the solid wall surface charge. Further, pmA and pmi are the porosities associated with macro- and micropores, respectively, defined per volume of the total electrode assembly (not, e.g., defined per carbon particle volume). We can further write a balance of charge equation as: pmi
∂cmA ∂(σmi2 /4 + [cmAe μatt ]2 )1/2 ∂c ⎞ ∂ ⎛ ⎜pmA D mA ⎟ + pmi = ∂t ∂t ∂x ⎝ ∂x ⎠
(6) 2010
dx.doi.org/10.1021/es403682n | Environ. Sci. Technol. 2014, 48, 2008−2015
Environmental Science & Technology
Article
light with peak intensity at 443 nm. SPQ fluorescence emissions are collisionally quenched by chloride ions (as well as other halide ions such as bromide and iodide 29), and its fluorescence intensity in a chloride ion containing salt solution (in the absence of other quenchers) can be described by a Stern−Volmer equation:40
subsequently etched into the carbon surfaces through the process of thermal activation (so-called “micropores”).38,39 In this work, we use HCAM electrodes which had been thermally activated (exposed to CO2 at 950 °C) for 2 h. The rest of the steps in fabrication of the HCAMs used in this study are as described in Baumann et al.38 HCAM electrodes have previously been utilized in flow-through electrode capacitive desalination systems8 and as electric double layer capacitor energy storage systems.39 We measured the total porosity (pmA + pmi) of our HCAM electrode as approximately 0.77 using dry/water-saturated measurements of the electrode mass. From previously reported N2 adsorption porosimetry measurements,27 we estimate that our electrode has a porosity associated with the micropores, pmi, of about 0.2, and so, we can approximate pmA = 0.77 − pmi = 0.57. In our CDI cell, the HCAM electrodes were separated by a distance of 1.5 mm. The step voltage applied to the cell, Vcell, varied between 0.25 and 1.25 V. Experimental Setup. In Figure 1b, we show a schematic of the experimental setup used for visualizations of salt concentration between two charging electrodes. In this system, we inserted the two HCAM electrodes into a transparent, 1.6 cm long, square borosilicate channel with a 5.9 × 5.9 mm internal cross section (Friedrich & Dimmock Inc., Millville, NJ, USA). Between the electrodes, we place a 5.8 × 5.8 mm acrylic frame with a thickness (along the x direction) of 1.5 mm (McMaster-Carr, Santa Fe Springs, CA, USA). The frame was three sided, with a slot of 5 × 5 × 1.5 mm laser cut into the open fourth side (Universal Laser Systems, Scottsdale, AZ, USA). The slot’s purpose was to allow for optical access into the space between the two HCAM electrodes. Two 5 × 5 mm, 127 μm thick titanium foil sheets (Alfa Aesar, Ward Hill, MA, USA) were inserted into the borosilicate channel and contacted with the outward-facing ends of the HCAM electrodes. Acrylic rectangles (so-called “holders” in Figure 1b) with cross-section dimensions of 5.8 × 5.8 mm and length (along the x direction) of 6.1 mm were placed behind the titanium foil, and a platinum wire was inserted through a hole drilled into the center of the “holder” to contact the foil. The platinum wire was used contact the external power supply (Keithley 2410, Cleveland, OH, USA) to the titanium foil current collector. On each side of the glass channel, two 1.5 mm thick acrylic end plates were bolted together, sandwiching a compressible expanded Teflon gasket, in order to seal the setup and compress the setup components for good electrical contact. Both end plates had external dimensions of 2 × 2 cm, and one end plate had a 7.6 × 7.6 mm window laser cut into its center in order to slip it over the borosilicate channel’s outer walls. For all visualizations, we used an inverted epifluorescent microscope (Eclipse TE300, Nikon, Japan), a 4×, NA 0.2 objective (Nikon, Japan), and captured images with a CCD camera (PI-MAX: 512, Princeton Instruments, Trenton, NJ, USA) with a 24 μm pixel width. Further, we used a 11000v3 Chroma filter cube (Chroma, Bellows Falls, VT, USA) with peak excitation wavelength range of roughly 320−380 nm and broadband emission beyond 420 nm. Chloride Ion Concentration Measurements. We measured spatially and temporally resolved salt concentration in the space between charging HCAM electrodes using species-altered fluorescence imaging (SAFI).29 As our fluorescent probe, we use 6-methoxy-N-(3-sulfopropyl)quinolinium (SPQ) (Santa Cruz Biotechnology Inc., Dallas, TX, USA). SPQ is excited by UV light at about 344 nm and subsequently emits visible
Io = 1 + KQ , cl−[ccl−] I
(12)
where Io is the fluorescence intensity at zero chloride ion concentration, I is the intensity at a chloride ion concentration of ccl−, and KQ,cl− is the quenching constant, which is 107 M−1 for a salt solution with solely chloride as the anions.41 A key feature of SPQ is that it is neutrally charged and so does not electromigrate despite the presence of an electric field in our charging CDI cell. 29 Thus, changes in observed SPQ fluorescence can be attributed to changing chloride ion concentration. We related measured SPQ fluorescence to local chloride ion concentration with the following equation, derived from eq 12: [ccl−] =
⎛ Iref ⎞ ⎜ (1 + KQ , cl−[ccl−, ref ]) − 1⎟ KQ , cl− ⎝ I ⎠ 1
(13)
Iref is the fluorescence intensity measured at a reference concentration, ccl−,ref. In our experiments, the reference fluorescence is that of the initial (undesalted) solution before application of the cell voltage. Experimental Protocol. In all experiments, we used an electrolyte consisting of a mixture of KCl salt and SPQ probe. We used KCl because the ionic mobilities of the K+ and Cl− ions are approximately equal, and thus, our experiments more closely adhered to the assumptions used in the porous electrode transport model (see Theory section). We used KCl solutions with concentrations of either 50 or 80 mM and an SPQ concentration of 5 mM in all experiments. The following protocol was used in all experimental results presented in this work. For at least 12 h before experiments, each HCAM electrode was soaked in a 10 mL solution of either 50 or 80 mM KCl and 5 mM SPQ (where each electrode held
